Calculation
On these pages we will publish documentation concerning compute calculations, high performance computing, big memory, etc. It is far from complete yet. For more questions, please contact our system administrators.
Overview of possibilities for doing calculation work at the MPI
- Your own desktop
- Windows
- Linux
- Department-specific Linux calculation servers
- LAG:
- Lux13 (16 cores, 132 GB RAM, OpenSuSE 15.2)
- Lux14 (8 cores, 284 GB RAM, OpenSuSE 15.2)
- NBL:
- Lux12 (24 cores, 512 GB RAM, OpenSuSE 15.2)
- LAG:
The MPI maintains its own compute cluster. This cluster is suited for highly parallel jobs which are processed in a batches and distributed among the different calculation nodes. Using the cluster requires UNIX skills, some scripting and a job suitable for the cluster. Contact the TG to discuss your job's suitability in advance and to get access permissions to the cluster.
There is a gridusers mailing list you can subscribe to in order to receive status updates and for general discussion.
- MPI Grid Engine
- Twelve nodes (20 cores, 256 GB RAM, OpenSuSE 15.2)
- One node (96 cores, 4 TB RAM, OpenSuSE 15.2)
- Fortytwo nodes (24 cores, 256 GB RAM, OpenSuSE 15.2) with:
- Quadro RTX 8000 (4608 cuda cores, 576 tensor cores, 48 GB GDDR6 RAM)
- Grid Engine description and documentation
- Parallel cluster storage 415 TB on /data/clusterfs
- Son of Grid Engine documentation: SGE Howto's
- Offsite calculation clusters
- RZG Garching
- SURFsara